Les microscopies optiques super-résolues aident à mieux comprendre certains mécanismes biomoléculaires, notamment au sein des neurones. Nous avons construit un tel microscope de type STED (STimulated Emission Depletion) pour observer des défauts azote-lacune (NV) fluorescents dans le diamant, et avons atteint une résolution de 50 nm. À plus long terme, cet instrument permettra d’étudier l’organisation macromoléculaire de protéines impliquées dans la plasticité synaptique, et marquées avec des nanodiamants (ND) fluorescents. Dans cette perspective, nous avons étudié la limite de résolution du STED pour des ND de tailles sub-longueur d’onde. Nos expériences, menées avec l’équipe de Stefan Hell (Max Planck Institute for Biophysical Chemistry) ont montré que la taille du spot STED d’un NV dans un ND pouvait atteindre 10 nm, performance similaire à celle obtenue dans un diamant macroscopique. Nous pouvons aussi résoudre plusieurs centres NV dans un ND séparés de seulement ~15 nm. Ces résultats sont en accord avec les simulations numériques faites par l’équipe de Jean-Jacques Greffet (Laboratoire Charles Fabry). En parallèle, nous avons démontré l’internalisation spontanée de ND fluorescents dans des neurones corticaux d’embryons de souris en culture primaire, et étudié leur colocalisation avec des vésicules du réseau trans-Golgi. Enfin, nous avons débuté l’étude du trafic des vésicules contenant les ND et montré qu’il dépend du réseau de microtubules. Les paramètres du mouvement sont compatibles avec ceux des moteurs moléculaires, mais nous nous attendons à ce qu’ils soient différents dans le cas de la surexpression de protéines impliquées dans le trafic (travail en cours). / Super resolution microscopy techniques are a useful tool to understand some biomolecular mechanisms, particularly in neurons. We have built such a STED (STimulated Emission Depletion) microscope for observing Nitrogen-Vacancy (NV) fluorescent defect in diamond, and have reached a resolution of 50 nm. In the longer term, this instrument will study the macromolecular organization of proteins involved in synaptic plasticity and marked with fluorescent nanodiamonds (ND). In this context, we studied the resolution limit of STED for ND of subwavelength size. Our experiments, conducted with the team of Stefan Hell (Max Planck Institute for Biophysical Chemistry) showed that the STED spot size of an NV in ND could reach 10 nm, which is similar to performances obtained in a macroscopic diamond. We can also resolve several NV centers, which are separated from only ~15 nm in the same ND. These results are in agreement with numerical simulations carried out by the team of Jean-Jacques Greffet (Laboratoire Charles Fabry). In parallel, we have demonstrated the spontaneous internalization of fluorescent ND in primary culture of cortical neurons from mouse embryos, and studied their colocalization with vesicles of the trans-Golgi network. Finally, we started the study of trafficking vesicles containing ND and showed that it depends on the microtubule network. The motion parameters are compatible with those of molecular motors, but we expect them to be different in the case of overexpression of proteins involved in traffic (work in progress).

La microscopie à éclairements structurés (SIM) permet théoriquement de doubler la résolution d’un microscope optique standard. Pour atteindre cette limite théorique, le SIM requière un contrôle très précis des illuminations, ce qui le rend coûteux et difficile à calibrer. Cette thèse cherche à simplifier drastiquement le principe du SIM en proposant une approche « aveugle » qui reconstruit une image de l’échantillon à partir d’éclairements aléatoires, i.e., très facile à générer. Cette stratégie permet en théorie l’imagerie super-résolue tout en réduisant fortement le coût de l’instrument. Nous avons étudié du point de vu théorique et algorithmique les performances et les limitations d’un estimateur joint de l’objet et des illuminations (estimateur Blind-SIM joint). Notamment, une reformulation mathématique du problème d’estimation jointe a été proposée qui permet d’analyser l'origine de la super-résolution mais également de proposer des nouvelles stratégies de mises en œuvre très rapides. Une étude empirique a mis en lumière l’impact de la parcimonie et du contenue fréquentielles des illuminations sur le niveau de super-résolution obtenu. L’estimateur joint étant asymptotiquement inconsistant, nous nous sommes également intéressé à définir un « critère de contraste » pour ce problème permettant d’estimer uniquement l’objet d’intérêt. Une étude mathématique de la capacité de super-résolution de ce type d'estimateur a été conduite. Enfin, on a observé un effet de super-résolution en condition réelles sur de nombreux objets, 2 ou 3D, fixe ou mobile, biologique ou non tel que des billes, des podosomes, de l’actine. / Structured illumination microscopy (SIM) allow theoretically to double the super-resolution of a standard optical microscope. However, to reach this theoretical limit, SIM require a precise knowledge of the illuminations, making it costly and difficult to calibrate. The aim of this thesis is to simplify the use of SIM by using a blind approach who allow the use of random illuminations to reconstruct a super-resolved image of the object. This strategy theoretically allow the super-resolution, while maintaining a low cost instrumentation. During those three years of thesis, we have studied theoretically and algorithmically the performances and the limitations of a joint estimator of the objet and the illuminations (joint Blind-SIM estimator). A mathematically equivalent reformulation of the joint problem was proposed allowing us to study the super-resolution origin and to propose a fast and parallelizable new approach. An empirical study has highlighted the impact of parsimony and of the frequency content of the illuminations on the reached super-resolution level. Because the joint estimator is asymptotically not consistent, we also studied a contrast criterion for our problem (typically a marginal likelihood), here only the object of interest is estimated. We have mathematically studied the super-resolution capacity of this kind of estimators. Finally, real data using random illuminations where acquired and we have observed a super-resolution effect using our algorithms on multiples real objects of different kind, 2 or 3D, fix or mobile, biological or not, like beads, podosomes, actines.

Num cenário em que dispositivos de aquisição de imagens e vídeo possuem recursos limitados ou as imagens disponíveis não possuem boa qualidade, as técnicas de super-resolução (SR) apresentam uma excelente alternativa para melhorar a qualidade das imagens. Nesta tese é apresentada uma proposta para super-resolução de imagem única que combina os benefícios da interpolação no domínio da transformada DCT e a eficiência dos métodos de reconstrução baseados no conceito de representação esparsa de sinais. A proposta busca aproveitar as melhorias já alcançadas na qualidade e eficiência computacional dos principais algoritmos de super-resolução existentes. O método de super-resolução proposto implementa algumas melhorias nas etapas de treinamento e reconstrução da imagem final. Na etapa de treinamento foi incluída uma nova etapa de extração de características utilizando técnicas de aguçamento por máscara de nitidez e construção de um novo dicionário. Esta estratégia busca extrair mais informações estruturais dos fragmentos de baixa e alta resolução do conjunto de treinamento e ao mesmo tempo reduzir o tamanho dos dicionários. Outra importante contribuição foi a inclusão de um processo iterativo e escalonável no algoritmo, reinserindo no conjunto de treinamento e na etapa de reconstrução, uma imagem de alta resolução obtida numa primeira iteração. Esta solução possibilitou uma melhora na qualidade da imagem de alta resolução final utilizando poucas imagens no conjunto de treinamento. As simulações computacionais demonstraram a capacidade do método proposto em produzir imagens com qualidade e com tempo computacional reduzido. / In a scenario in which the acquisition systems have limited resources or available images do not have good quality, the super-resolution (SR) techniques have become an excellent alternative for improving the image quality. In this thesis, we propose a single-image super-resolution (SR) method that combines the benefits of the DCT interpolation and efficiency of sparse representation method for image reconstruction. Also, the proposed method seeks to take advantage of the improvements already achieved in quality and computational efficiency of the existing SR algorithms. The proposed method implements some improvements in the dictionary training and the reconstruction process. A new dictionary was built by using an unsharp mask technique to characteristics extraction. Simultaneously, this strategy aim to extract more structural information of the low resolution and high resolution patches and reduce the dictionaries size. Another important contribution was the inclusion of an iterative and scalable process by reinserting the HR image obtained of first iteration. This solution aim to improve the quality of the final HR image using a few images in the training set. The results have demonstrated the ability of the proposed method to produce high quality images with reduced computational time.

L'objectif de cette thèse est d'étudier le comportement de différentes représentations d'images, notamment apprentissage profond, dans le contexte d'application en imagerie médicale. Le but est de développer une méthode unifiée efficace pour les applications visées que sont la super résolution, la segmentation et la synthèse. La super-résolution est un procès d'estimation d'une image haute-résolution à partir d'une ou plusieurs images basses résolutions. Dans cette thèse, nous nous concentrons sur la super résolutionunique, c'est-à-dire que l'image haute résolution (HR) est estimée par une image basse-résolution (LR) correspondante. Augmenter la résolution de l'image grâce à la super-résolution est la clé d'une compréhension plus précise de l'anatomie. L'application de la super résolution permet d'obtenir des cartes de segmentation plus précises. Étant donné que deux bases de données qui contiennent les images différentes (par exemple, les images d'IRM et les images de CT), la synthèse est un procès d'estimation d'une image qui est approximative aux images dans la base de données de cible à partir d'une image de la base de données de source. Parfois, certains contrastes tissulaires ne peuvent pas être acquis pendant la séance d'imagerie en raison du temps et des coûts élevés ou de l'absence d'appareils. Une solution possible est à utiliser des méthodes de synthèse d'images médicales pour générer les images avec le contraste différent qui est manquée dans le domaine à cible à partir de l'image du domaine donnée. L'objectif des images synthétiques est d'améliorer d'autres étapes du traitement automatique des images médicales telles que la segmentation, la super-résolution ou l'enregistrement. Dans cette thèse, nous proposons les réseaux neurones pour la super résolutionet la synthèse d'image médicale. Les résultats démontrent le potentiel de la méthode que nous proposons en ce qui concerne les applications médicales pratiques. / In this thesis, our motivation is dedicated to studying the behaviors of different image representations and developing a method for super-resolution, cross-modal synthesis and segmentation of medical imaging. Super-Resolution aims to enhance the image resolution using single or multiple data acquisitions. In this work, we focus on single image super-resolution (SR) that estimates the high-resolution (HR) image from one corresponding low-resolution (LR) image. Increasing image resolution through SR is a key to more accurate understanding of the anatomy. The applications of super-resolution have been shown that applying super-resolution techniques leads to more accurate segmentation maps. Sometimes, certain tissue contrasts may not be acquired during the imaging session because of time-consuming, expensive costor lacking of devices. One possible solution is to use medical image cross-modal synthesis methods to generate the missing subject-specific scans in the desired target domain from the given source image domain. The objective of synthetic images is to improve other automatic medical image processing steps such as segmentation, super-resolution or registration. In this thesis, convolutional neural networks are applied to super-resolution and cross-modal synthesis in the context of supervised learning. In addition, an attempt to apply generative adversarial networks for unpaired cross-modal synthesis brain MRI is described. Results demonstrate the potential of deep learning methods with respect to practical medical applications.

Single-molecule localisation microscopy (SMLM), has allowed for optical microscopy to probe biological systems beyond the diffraction limit. The intrinsic 3D nature of biology has motivated the development of 3D-SMLM with novel techniques, including the double-helix point spread function (DHPSF). A bespoke microscope platform employing the DHPSF transformation was built, achieving ~10 nm lateral and ~20 nm axial localisation precision over a ~4 μm axial depth. Until recently, the DHPSF has been limited by spherical aberration present when imaging away from coverslip surfaces to the study of small volumes close to the coverslip. By matching the refractive index of the objective lens immersion liquid to that of the imaging media, this aberration can be minimised, facilitating large-volume imaging away from unphysiological flat surfaces. The work presented in this thesis illustrates the capabilities of the DHPSF for 3D-SMLM and single-particle tracking (SPT) in previously inaccessible areas of biological samples (e.g. in the nucleus and on the apical cell surface). Application of the DHPSF for SPT in eukaryotic cells are presented; tracking the motion of T-cell membrane proteins on the apical surface and components of the chromosome remodelling complex in the nucleus of embryonic stem cells. For these applications, meansquared displacement and jump distance diffusion analysis methodologies were extended into 3D and benchmarked against simulated datasets. A variety imaging applications that are facilitated by the extended depth of focus of the DHPSF are presented, focusing on quantification of T-cell membrane protein reorganisation upon immunological activation. Finally, the clustering distribution of the T-cell receptor is investigated by Ripley’s K analysis enabled by duel labelling of its position and the outer membrane in primary T cells.

Recently, there are great interests in the location-based applications and the location-awareness of mobile wireless systems in indoor areas, which require accurate location estimation in indoor environments. The traditional geolocation systems such as the GPS are not designed for indoor applications, and cannot provide accurate location estimation in indoor environments. Therefore, there is a need for new location finding techniques and systems for indoor geolocation applications. In this thesis, a wide variety of technical aspects and challenging issues involved in the design and performance evaluation of indoor geolocation systems are presented first. Then the TOA estimation techniques are studied in details for use in indoor multipath channels, including the maximum-likelihood technique, the MUSIC super-resolution technique, and diversity techniques as well as various issues involved in the practical implementation. It is shown that due to the complexity of indoor radio propagation channels, dramatically large estimation errors may occur with the traditional techniques, and the super-resolution techniques can significantly improve the performance of the TOA estimation in indoor environments. Also, diversity techniques, especially the frequency-diversity with the CMDCS, can further improve the performance of the super-resolution techniques.

National security needs dictate the development of new radar systems capable of identifying and tracking exoatmospheric threats to aid our defense. These new radar systems feature reduced noise floors, electronic beam steering, and ultra-wide bandwidths, all of which facilitate threat discrimination. However, in order to identify missile attributes such as RF reflectivity, distance, and velocity, many existing processing algorithms rely upon narrow bandwidth assumptions that break down with increased signal bandwidth. We present a fresh investigation into these algorithms for removing bandwidth limitations and propose novel state-space and direct-data factoring formulations such as * the multidimensional extension to the Eigensystem Realization Algorithm, * employing state-space models in place of interpolation to obtain a form which admits a separation and isolation of solution components, * and side-stepping the joint diagonalization of state transition matrices, which commonly plagues methods like multidimensional ESPRIT. We then benchmark our approaches and relate the outcomes to the Cramer-Rao bound for the case of one and two adjacent reflectors to validate their conceptual design and identify those techniques that compare favorably to or improve upon existing practices.

Amyloidogenic neurodegenerative disorders such as Alzheimer's disease (AD) and Parkinson's disease (PD) are an important health issue. However, the underlying molecular mechanisms of the disease-related protein aggregates, that are present in humans, are only understood partially. I have used and developed biophysical methods to study the structural and biological properties of individual aggregates of Amyloid β peptide and α-Synuclein, proteins whose aggregation is associated with the development of Alzheimer's and Parkinson's disease respectively. I expanded the single aggregate visualisation through enhancement (SAVE) technique, which is a method based on the fluorescent dye Thioflavin T (ThT) that reversibly bind to the aggregates and whose fluorescence increases upon binding. I firstly explored the use of other dyes for these experiments and found that a ThT dimer has higher affinity to α-Synuclein aggregates in vitro. I then applied the SAVE method to the cerebral spinal fluid (CSF) of a cohort of AD patients and control CSF and observed no clear difference in aggregate number. However, these experiments provided insights into how antibodies bind the aggregates in human CSF. I could show, that despite altering the Ca2+ influx into both cells and vesicles, the antibody did not measurably affect the aggregate structure. To study the size specific effects of the Amyloid β 42 (Aβ42) peptide in more detail, I used and optimised gradient ultracentrifugation combined with single aggregate imaging to study the structural properties of the isolated aggregates. This aggregation kinetic independent method allowed me to compare the properties of fluorescently labelled and unlabelled Aβ42 and characterize the size dependent properties of aggregates in a single experiment. Since I could measure the relative concentration of different size aggregates it was also possible to compare the properties of single aggregates of different sizes. I then used biological assays to examine the ability of aggregates to permeabilise membranes resulting in the entry of calcium ions, and their ability to induce TNFα production in microglia cells. Both processes are thought to play key roles in the development of AD. I found that small soluble oligomers are most potent at inducing Ca2+ influx, whereas longer protofilaments are the most potent inducers of TNFα production. My results suggest that the mechanism by which aggregates damage cells changes as aggregation proceeds, as longer aggregates with different structures are formed. Protofilaments with a diameter of 1 nm or less have a structure that could make them particularly potent at causing the signalling of toll-like receptors, providing a molecular basis for their ability to induce TNFα production.

Des techniques variées de visualisation de molécules d’intérêt sur cellules vivantes ou fixées ont permis de suivre leur synthèse, localisation, dégradation et autres activités. Dans cette étude, nous avons développé deux outils de fluorescence pour étudier la synthèse des protéines sur bactéries vivantes. Le premier décrit l’utilisation du système Spinach pour l’imagerie du ribosome. Cette approche diffère des méthodes conventionnelles qui utilisent des protéines fluorescentes puisque l’ARN ribosomal 16S contient un aptamère qui rend fluorescent un composé fluorogène. Une étude comparative de la performance de différents aptamères Spinach a été réalisée. Un deuxième outil se focalise sur l’accumulation d’un antibiotique de la famille des aminoglycosides (ligand du ribosome) conjugué à un fluorophore. Ce nouveau conjugué, qui a conservé son activité bactéricide permet pour la première fois de visualiser l’accumulation de l’antibiotique sur bactérie vivante. Cela permet une analyse au niveau de la cellule unique d’une population bactérienne exposée à l’antibiotique. Nous avons également obtenu des données sur la localisation de l’antibiotique une fois qu’il a pénétré dans la bactérie à une résolution inégalée par microscopie super-résolutive. Nous espérons que ces deux méthodes vont maintenant permettre une meilleure compréhension de la synthèse des protéines et fournir une vue nouvelle de la pénétration des antibiotiques dans les bactéries pour y produire leur action bactéricide. / Various visualizing techniques have previously enabled monitoring the fate of molecules of interest: their expression, localization, degradation and other activities in live or fixed cells. In this study, we have developed two fluorescent tools to study protein synthesis in live bacterial cell. The first one describes the application of Spinach system to ribosomes imaging. This is different from conventional methods (that use fluorescent proteins) in that 16S rRNA contains an inserted RNA aptamer that elicits fluorescence of a fluorogenic compound. A comparative study of the performance of different Spinach aptamers was performed here. A second system focuses on the uptake of a fluorescently labeled ligand of the ribosome, an antibiotic of the class of aminoglycosides. This novel conjugate, which kept its bactericidal activity allows for the first time imaging of aminoglycoside uptake on live bacteria. This opened the door to a single cell analysis of bacterial cell populations. We also obtained data about the localization of the antibiotic once inside the bacteria to an unprecedented resolution using super resolution microscopy. We hope that both of these methods will contribute to a better understanding of protein synthesis as well as provide a novel view on the way antibiotics penetrate into cells and perform their bactericidal action.

Ce travail de thèse concerne l'amélioration du pouvoir de résolution de la microscopie optique en champ lointain. Nous avons développé des techniques qui tirent profit de la relation liant le champ électromagnétique émis par un objet à l’éclairement utilisé. En utilisant plusieurs images obtenues pour différents éclairements, et à l’aide d’un algorithme d'inversion approprié, il est possible d'accéder à des fréquences spatiales de l'objet habituellement filtrées par le microscope.Ce concept est d’abord appliqué à une technique de microscopie cohérente : la tomographique optique de diffraction. Elle permet d’obtenir numériquement une carte quantitative de la permittivité diélectrique de l'objet, avec une résolution supérieure à celle d'un microscope classique, à partir de plusieurs hologrammes de l'échantillon. Dans ce cadre, nous montrons que le phénomène de diffusion multiple permet d’atteindre des résolutions encore plus spectaculaires s’il est pris en compte. Nous étudions ensuite la microscopie de fluorescence par éclairement structuré, que nous proposons d’améliorer de deux manières différentes. Dans la première, nous utilisons un algorithme d’inversion capable de retrouver simultanément la densité de fluorescence et les éclairements utilisés. Grâce à celui-ci, nous pouvons remplacer l’illumination périodique et contrôlée généralement utilisée, par des speckles aléatoires formés avec un montage remarquablement simple. Nous montrons expérimentalement l'efficacité de cette approche. Dans un second temps, nous proposons de remplacer la lamelle de verre sur laquelle est repose l’échantillon par un réseau diélectrique nanométrique. Celui-ci crée à sa surface une grille de lumière de période inférieure à la limite de diffraction, ce qui permet d’améliorer d’avantage la résolution finale de l’image reconstruite. Nous détaillons la conception, la fabrication et la caractérisation expérimentale de ce substrat nanostructuré. / This PhD work focuses on the resolution improvement of far-field optical microscopy. We have studied and developed different techniques that take advantage of the relationship between the sample, the illumination and the diffracted (or emitted) field, in order to increase final band-pass of the image beyond that imposed by the diffraction phenomenon. In In these approaches, several images of the same sample are recorded under different illuminations. An inversion algorithm in then used to reconstruct a super-resolved map of the sample from the set of measurements.This concept is first applied to coherent microscopy. In tomographic diffraction microscopy, many holograms of the same unstained sample are obtained under various incidences, then used to numerically reconstruct a quantitative map of permittivity of the sample. The resolution is usually better than that of classical wide-field microscopy. We show theoretically and experimentally that, far from being a drawback, the presence of multiple scattering within the sample can, if properly accounted for, lead a to an even better resolution.We then study structured illumination fluorescence microscopy. We present two different ways for improving this method. The first one takes advantage of an inversion algorithm, which is able to retrieve the fluorescence density without knowing the illumination patterns. This algorithm permits one to replace the periodic light pattern classically used in structured illumination microscopy by unknown random speckle patterns. The implementation of the technique is thus considerably simplified while the resolution improvement remains. In the second approach, we propose to replace the coverslip on which the sample usually lays, by a sub-lambda grating. The latter is used to form, in near field, a light grid with sub-diffraction period that is able to probe the finest details of the sample. The design, fabrication and optical characterization of this key structure are detailed.